1. Field of the Invention
The present invention relates to the detection of cryptographic attacks on cryptographic calculation units, and particularly to detection of influencing the cryptographic units from outside within the scope of the cryptographic attack.
2. Description of the Related Art
Two of the best know attack scenarios on security devices are the differential power analysis (DPA) and fault attacks (FA). Every kind of these devices has to be capable of fending off these attacks.
In a DPA attack, information leading, for example, to a disclosure of the private key used by the cryptographic calculation is derived on the basis of a measurement of power consumed by a cryptographic device during a cryptographic calculation with predetermined attack parameters.
In order to thwart a DPA attack, different time delays may be implemented to change the processing patterns taken advantage of by a DPA attack, so that the characteristic features necessary for a disclosure of the private key do not occur in the power course.
Another possibility is to mask the cryptographic data, wherein the cryptographic data are combined with masking numbers, for example, such that power analysis does not lead to the desired success.
The fault attacks, however, are cryptographic attacks in which, for example, by outside effect on cryptographic chip, faults in the cryptographic calculation are brought about, which may lead to a disclosure of the private key.
Fault attacks may, for example, be fended off with sensor technology and special software. But the first only solves the problem symptomatically, since an influence, but not the effect, is recognized. The latter is only possible in special situations, for example in an RSA calculation, but not so much in the general execution of the code in the CPU.
DPA attacks may, for example, be fended off by generation of a balanced current profile, e.g. by dual rail with pre-charge in full custom, by generation of current or time noise, or by randomization of the process data. Again, the latter is only possible in special situations through the software, for example for the RSA calculation. The first, however, demands very intensive design, so that a semi-custom design is not yet possible, and does not necessarily guarantee success. Noise as masking parameter may complicate DPA analysis, but does not render it impossible.